U.S. patent application number 14/251312 was filed with the patent office on 2014-10-16 for inspecting high-resolution photolithography masks.
This patent application is currently assigned to KLA-Tencor Corporation. The applicant listed for this patent is KLA-Tencor Corporation. Invention is credited to David Alles, Gregg Anthony Inderhees, Mehdi Vaez-Iravani, Fred Stanke, Stanley E. Stokowski, Ilya Toytman.
Application Number | 20140307943 14/251312 |
Document ID | / |
Family ID | 51686845 |
Filed Date | 2014-10-16 |
United States Patent
Application |
20140307943 |
Kind Code |
A1 |
Stanke; Fred ; et
al. |
October 16, 2014 |
INSPECTING HIGH-RESOLUTION PHOTOLITHOGRAPHY MASKS
Abstract
Optical inspection methods and apparatus for high-resolution
photomasks using only a test image. A filter is applied to an image
signal received from radiation that is transmitted by or reflected
from a photomask having a test image. The filter may be implemented
using programmed control to adjust and control filter conditions,
illumination conditions, and magnification conditions.
Inventors: |
Stanke; Fred; (San Jose,
CA) ; Toytman; Ilya; (Menlo Park, CA) ; Alles;
David; (Los Altos, CA) ; Inderhees; Gregg
Anthony; (Morgan Hill, CA) ; Stokowski; Stanley
E.; (Danville, CA) ; Mehdi Vaez-Iravani;; (Los
Gatos, CA) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
KLA-Tencor Corporation |
Milpitas |
CA |
US |
|
|
Assignee: |
KLA-Tencor Corporation
Milpitas
CA
|
Family ID: |
51686845 |
Appl. No.: |
14/251312 |
Filed: |
April 11, 2014 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
61812248 |
Apr 16, 2013 |
|
|
|
Current U.S.
Class: |
382/144 ;
348/92 |
Current CPC
Class: |
G01N 2021/95676
20130101; G06T 2207/10056 20130101; G06T 2207/30148 20130101; G06K
9/00 20130101; G06T 7/001 20130101; G01N 21/956 20130101 |
Class at
Publication: |
382/144 ;
348/92 |
International
Class: |
G06T 7/00 20060101
G06T007/00; G01N 21/95 20060101 G01N021/95 |
Claims
1. A method comprising: illuminating a test pattern on a
high-resolution lithographic mask with an incident light beam
having a characteristic optical bandwidth and focused through an
optical element; capturing light that is transmitted through the
mask or reflected by the mask, the captured light forming a sensed
image file; and using a processor to: apply a filter function to
the sensed image file to obtain a filtered image file having
desired frequency characteristics; perform a mathematical operation
in the frequency domain using the sensed image file and the
filtered image file to obtain a resultant image in accord with an
image processing scheme; and evaluate the resultant image for
indications of defects in the lithographic mask.
2. The method of claim 1, wherein the illuminating step includes
illuminating the test pattern with a coherent incident light
beam.
3. The method of claim 2, wherein the coherence of the incident
light beam is adjustable, and the processor is used to adjust the
coherence.
4. The method of claim 1, wherein the optical element has a
numerical aperture, the method further comprising restricting the
numerical aperture such that the test pattern falls outside the
characteristic optical bandwidth of the incident light beam.
5. The method of claim 4, wherein the numerical aperture of the
optical element is adjustable, and the processor is used to adjust
the numerical aperture.
6. The method of claim 1, wherein the step of capturing light uses
time delay and integration based imaging.
7. The method of claim 1, wherein the step of applying a filter
function includes applying a Fourier Transform to the sensed image
file.
8. The method of claim 1, wherein the step of applying a filter
function includes filtering out all images in the sensed image file
not having the desired frequency characteristics, and subtracting
the filtered sensed image from the sensed image file to form the
resultant image.
9. The method of claim 1, wherein the step of capturing light
includes scanning the test pattern in a lateral dimension relative
to a plane of the mask.
10. The method of claim 9, wherein the lateral dimension is
arranged to be parallel with linear features formed on the test
pattern.
11. The method of claim 10, further comprising: scanning the test
pattern in the X dimension to return an array of image intensity
values in the Y dimension; and subtracting an average of the image
intensity values in the Y dimension from the sensed image to form a
resultant image.
12. A method, comprising: illuminating a test pattern on a
high-resolution lithographic mask with an incident light beam
having a characteristic optical bandwidth and focused through an
optical element having a characteristic numerical aperture;
optically filtering light that is transmitted through the mask or
reflected by the mask to obtain a resultant image having desired
frequency characteristics; capturing the resultant image with a
sensor; and evaluating the resultant image for indications of
defects in the lithographic mask.
13. The method of claim 12, wherein the illuminating step is
controlled by a processor that is configured to adjust the optical
bandwidth of the incident light beam and to adjust the numerical
aperture of the optical element.
14. The method of claim 13, wherein the illuminating step includes
illuminating the test pattern with a coherent incident light beam
having an adjustable coherence, and the processor is configured to
adjust the coherence of the incident light beam.
15. The method of claim 12, wherein the filtering step is
controlled by a processor that is configured to perform a filtering
function on the sensed image file to obtain a filtered image file
having desired frequency characteristics, and to perform a
mathematical operation in the frequency domain using the sensed
image file and the filtered image file to obtain the resultant
image in accord with an image processing scheme.
16. The method of claim 12, wherein the step of filtering includes
filtering out all images in the sensed image file not having the
desired frequency characteristics, and subtracting the filtered
sensed image from the sensed image file to form the resultant
image.
17. An optical inspection apparatus comprising: a stage for holding
a high-resolution lithographic mask; a light source configured to
emit an incident light beam toward the mask along an optical axis,
the incident light beam having a characteristic optical bandwidth;
a first optical element having a characteristic numerical aperture
and positioned on the optical axis between the light source and the
mask and configured to focus the incident light beam onto a test
image formed on the mask; a sensor configured to capture light that
is transmitted through the mask or reflected by the mask, the
captured light forming a sensed image file; an optical filter
coupled to receive the sensed image and configured to obtain a
filtered image having desired frequency characteristics; and a
processing element configured to perform a mathematical operation
in the frequency domain using the sensed image file and the
filtered image file to obtain a resultant image in accord with an
image processing scheme, and to evaluate the resultant image for
indications of defects in the lithographic mask.
18. The optical inspection apparatus of claim 17, wherein the
processing element is configured to adjust the optical bandwidth of
the incident light beam and to adjust the numerical aperture of the
optical element.
19. The optical inspection apparatus of claim 18, wherein the light
source emits a coherent incident light beam having an adjustable
coherence, and the processor is configured to adjust the coherence
of the incident light beam.
20. The optical inspection apparatus of claim 12, wherein the
optical filter filters out all images in the sensed image file not
having the desired frequency characteristics, and the processing
element subtracts the filtered sensed image from the sensed image
file to form the resultant image.
Description
TECHNICAL FIELD
[0001] This disclosure relates generally to semiconductor device
fabrication using photolithographic masks, and in particular, to
the inspection of high-resolution photolithographic masks.
BACKGROUND
[0002] The fabrication of semiconductor devices involves processing
a substrate, such as a semiconductor wafer, to form various
integrated circuit features on the substrate. One of the
fundamental steps of any semiconductor fabrication process is
photolithography, which is a series of steps for building layers of
a three-dimensional circuit structure on the substrate. In each
photolithography step, a light sensitive photoresist material is
applied to the substrate through a mask, exposed to a light source,
developed and then etched to form a portion of the
three-dimensional structure.
[0003] The patterns formed during lithography directly affect the
viability and fidelity of the intended integrated circuit features
that are ultimately formed. While particles and defects are
undesirable at any stage of the fabrication process, mask defects
are particularly problematic since they will affect many different
devices. Consequently, any defects formed as a result of
lithography, such as the transfer of defects that are present on
the mask, are problematic for the integrated circuit manufacturing
process. Thus, inspection of masks for defects is an important part
of quality control for the manufacturing process.
[0004] Fabrication processes typically include an inspection tool
for inspecting masks for defects in order to predict whether the
projected pattern image will faithfully reproduce the intended
design of the device. In general, an inspection tool images a test
image and a reference image on the mask and detects defects from
processing the images. In many cases, an optical inspection
apparatus does a good job of resolving the desired patterns on the
mask, and sophisticated detection methods may be employed to
separate the desired features of the image from defects that are
undesirable.
[0005] However, for some lithographic methods, the masks are formed
with high-resolution patterns that are not resolved well or perhaps
not at all by the optical inspection apparatus, and therefore
defect detection can be difficult in these cases. For example, an
optical inspector having a numerical aperture ("NA") of 0.75 and an
inspection wavelength of 200 nm has a "nominal resolution" of 130
nm for 4.times. features on the mask, which is equivalent to 35 nm
resolution on a wafer. Thus, the inspector has a hard optical
cutoff for features having a half-pitch of 19 nm on the wafer.
There is current interest in extreme ultraviolet ("EUV")
lithography to produce, for example, half-pitch lines at 16 nm,
which would be completely invisible to the inspector described
above. Further, efforts to produce even finer half-pitch lines with
nano-imprint lithograph ("NIL") are underway, for example, down to
6 nm half pitch, which would also be invisible to the inspector
mentioned above.
[0006] Using conventional methods, at least two images are
compared, which requires significant computing power and memory to
choose or form the reference, to do subtractions, and to make the
images as similar as possible for processing. Also, subtraction of
images may result in the noise coming from both images combining to
form a total noise component which is larger for the difference
than for either of the contributors. Thus, it would be desirable to
have alternative methods for inspecting, particularly for high
resolution masks.
BRIEF DESCRIPTION OF THE DRAWINGS
[0007] FIG. 1A is a schematic diagram illustrating the contact
method of printing patterns from a lithographic mask.
[0008] FIG. 1B is a schematic diagram illustrating the proximity
method of printing patterns from a lithographic mask.
[0009] FIG. 1C is a schematic diagram illustrating the projection
method of printing patterns from a lithographic mask.
[0010] FIG. 2 is a schematic diagram illustrating a tool for
printing patterns from a lithographic mask onto a wafer.
[0011] FIG. 3 is a schematic diagram illustrating a tool for
inspecting a lithographic mask.
[0012] FIG. 4A is a flow chart illustrating a process for capturing
and filtering an image that is directed onto a lithographic
mask.
[0013] FIG. 4B is a flow chart illustrating an example of a
filtering process.
[0014] FIG. 5A is a simplified schematic diagram illustrating an
arrangement for filtering an image before it is directed onto a
lithographic mask.
[0015] FIG. 5B is a simplified schematic diagram illustrating an
arrangement for filtering an image after it is directed onto a
lithographic mask.
[0016] FIG. 5C is a simplified schematic diagram illustrating an
arrangement for filtering an image directed onto a lithographic
mask using a signal processor.
[0017] FIG. 5D is a simplified schematic diagram illustrating an
arrangement for filtering an image directed onto a lithographic
mask using a signal processor and for controlling the illumination
optics and magnification optics using the signal processor.
DETAILED DESCRIPTION
[0018] This disclosure describes embodiments of an optical
inspection apparatus that are inspecting high-resolution photomasks
using optimized optical conditions and without differencing test
and reference images.
[0019] 1. Prior Inspection Methods
[0020] Using conventional optical inspection methods, a reference
image is formed for each portion of the mask that is being
sequentially tested for fidelity, and the reference image is
compared to a test image. The optical inspection apparatus can
derive reference images in various ways. For example, another
portion of the same mask that has the same intended pattern can be
chosen as the test portion. If there are multiple identical dice on
the mask, the reference may be chosen from a different die. In this
case, whether the defect is actually in the test or the reference
portion is ambiguous. As is well known in the art, if there are
more than two dice on the mask, the inspector can use double
arbitration. In the case where circuits are made of repeating
cells, as is frequently the case, for example, with memory
circuits, the inspection apparatus can use a different cell as the
reference to compare to the test cell. Such "cell-to-cell"
inspection typically requires that the lateral, one- or
two-dimensional cell pitch to be known, in order to facilitate
finding the right portion of the mask to use as a reference. In yet
another case, the inspection apparatus can read a database of the
circuit design for the layer being inspected, use knowledge of its
own optical system, and render an expected optical image of the
test region as a reference.
[0021] When comparing a test image to the reference image, the
inspection apparatus can adjust parameters in order to make the
test and reference images as similar as possible within certain
constraints. For example, the inspection apparatus can be
configured to shift one or the other of the test and reference
images in order to align them. In another example, the inspection
apparatus can equalize the color scales of the images to minimize
the difference between them. Finally, the inspection apparatus can
detect defects by looking for deviations in the difference between
the test and reference images that exceed some threshold level. For
a portion of the mask that is defect-free, the difference between
test and reference image is expected to be nominally zero, with
some noise due to tool issues such as shot noise and speckle in the
images, changes in focus between dice or cells, and errors in the
rendering of the database design to an optical image. The
inspection apparatus can erroneously detect defects if the noise
exceeds the detection threshold, and such false detection is
obviously highly undesirable.
[0022] While typical photolithography methods use a 4:1 optical
demagnification of the mask onto the substrate, making the actual
structures on the mask quite large, high-resolution methods, such
as nano-imprint lithography ("NIL"), uses contact transference of
patterns on the mask with a demagnification factor of 1:1. In some
cases, for producing the same target structures on the substrate,
an optical inspection tool will be able to resolve the pattern on
an optical mask but not on the NIL mask.
[0023] For example, FIG. 1A illustrates the method for contact
transference of patterns, where a light beam 2 is directed through
illumination optics 10 onto mask 11, which is in direct contact
with the photoresist layer 12 on wafer 13. FIG. 1B illustrates the
method for proximity printing, where the light beam 2 is directed
through illumination optics 20 and onto mask 11, then across a
small gap (10 to 25 microns) to the photoresist layer 12 on wafer
13. Finally, FIG. 1C illustrates the method for projection
printing, where the light beam 2 is directed through illumination
optics 30 onto mask 11, then through magnification optics 40 to the
photoresist layer 12 on wafer 13.
[0024] One option for inspecting NIL masks is to use a
scanning-electron microscope ("SEM"), which produces images by
scanning a sample one portion at a time with a focused beam of
electrons. However, SEM scanning is generally much slower than
optical scanning, which collects the entire image simultaneously.
Thus, the loss of useable mask time due to SEM inspection is
typically cost prohibitive.
[0025] In another case, high-resolution patterns can also be formed
using extreme ultraviolet ("EUV") lithography to transfer patterns
from an EUV mask to the substrate using radiation having an
approximate wavelength of 13.5 nm. At the present time, however,
there are no mask inspection tools using EUV that can provide
comparable resolution in the same modes as optical inspectors
inspecting optical masks. An optical inspector operating at a
wavelength of .about.193 nm, for example, can be used to inspect
EUV masks, but may have little or no resolution of the desired
patterns on the masks.
[0026] In yet another case, hybrid lithography methods are being
developed due to the delay in development of practical EUV
lithography. For example, it is possible to produce very fine, one
dimensional lines and spaces on a mask in a single lithography
step, and then make low resolution cuts in the lines with another
step. This method places tight constraints on the device design,
which can be expensive to implement, but worthwhile in terms of
manufacturability.
[0027] Methods for processing test and reference images to
determine defects are known, for example, from U.S. Pat. No.
7,564,345 entitled Inspection Method and Systems for Lithographic
Masks, which is incorporated by reference in its entirety herein.
Methods and apparatus for inspecting lithographic masks are known
from U.S. Pat. No. 7,486,393 entitled Multiple Beam Inspection
Apparatus and Method, which is incorporated by reference in its
entirety herein.
[0028] 2. Optical Inspection Apparatus
[0029] FIG. 2 is a simplified schematic diagram of an apparatus 100
useful for transferring a pattern from a mask M onto a wafer W as
part of a semiconductor fabrication process, while FIG. 3 is a
simplified schematic diagram of an apparatus 200 useful for
inspecting the mask M prior to using the mask in a semiconductor
fabrication process.
[0030] Referring to FIG. 2, the lithographic pattern transfer
apparatus 100 may be implemented as a scanner or stepper, such as
the PAS 5500 automatic stepper available from ASML Holding NV.
Other similar machines are available from Nikon Corporation, Canon,
Inc. and Ultratech, Inc., for example.
[0031] In general, an illumination source 102 emits radiation 103
of a desired wavelength through illumination optics 110 onto one or
more selected portions of the mask M. In one simplified embodiment,
the illumination optics include a first lens arrangement 111 for
collimating the light beam 103 into collimated light beam 113; a
modulator 112 for restricting the transmission field; a second lens
arrangement 114 for focusing the beam 113 that is transmitted
through the modulator into one or more illuminations spots having
the specified numeric aperture NA-1 at the plane of the mask M; and
a third lens arrangement 115 for focusing the radiation that is
transmitted through the mask onto the substrate W to create the
desired pattern on the substrate.
[0032] Modern lithographic transfer and inspection systems are
known that provide desired radiation at wavelengths of 248 nm, 193
nm, 157 nm, or extreme ultra violet ("EUV") at 13 nm, or
nanoimprint lithography ("NIL") at less than 10 nm resolution, or
even electron-beam or particle-beam exposure.
[0033] Referring to FIG. 3, one embodiment of the inspection
apparatus 200 has an optics system 210 similar to that of FIG. 1,
with first lens 211, collimated beam 213, modulator 212, second
lens 214, and third lens 215. However, in this embodiment, lens 215
includes microscopic magnification optics designed to provide
60-350.times. magnification for enhanced inspection. Thus, the
numerical aperture NA-2 at the plane of the mask M-2 on the
inspector tool is significantly greater than the numerical aperture
NA-1 at the plane of mask M-1 of the lithography tool. Thus, each
of these optical systems induce different optical effects in the
images produced.
[0034] An embodiment of the present invention can utilize standard
inspection apparatus 200 such as described in simplified schematic
representation of FIG. 3. Such an embodiment uses an illumination
source 202 that produces light 203 that is directed through
illumination optics 210 to produce a light beam 213 that is
directed onto and through a portion of the mask M-2.
[0035] For example, the illumination source 202 can be a laser or a
filtered lamp. In one embodiment, the illumination source 202 is a
193 nm laser. As explained above, the inspection apparatus 200 is
configured with a numerical aperture NA-2 at the plane of mask M-2
that is higher than the numerical aperture NA-1 at the plane of
mask M-1 of the lithography tool. The mask M-2 to be inspected is
placed on a horizontal plane at a specified location and exposed to
the illumination source 202. The patterned image from the portion
of mask M-2 is directed through the microscopic magnification
optical system 215 which projects the patterned image onto an
electronic sensor 250.
[0036] The sensor 250 collects an image that is transmitted through
the mask, or alternatively, or in addition, an additional sensor
and associated optics (not shown) may be positioned to collect an
image that is reflected by the mask. In one embodiment, the sensor
250 is a sensor array using time delay integration ("TDI")
technology, although other types of detectors could be used,
including charged coupled device ("CCD") sensors, CCD arrays, TDI
sensors, PMTs, and other electronic sensors known to those with
skill in the art.
[0037] The images captured by the sensor 250 are processed by
processing circuitry 260. The processing circuitry 260 may form a
part of the inspection apparatus, or alternatively, can be a
separate apparatus located remotely from the inspection apparatus.
In one embodiment, the processing circuitry 260 includes one or
more microprocessors and associated memory embodied in a computer
apparatus and configured to enable the principles described herein.
For example, the Teron series Reticle Defect Inspection Systems,
made by KLA-Tencor Corporation, of San Jose, Calif., are
computer-controlled inspection systems that targets 1.times.nm
logic and 2.times.nm half-pitch memory nodes.
[0038] 3. Improved Optical Inspection Methods
[0039] However, for high-resolution masks, such as NIL or EUV
masks, less expensive methods of inspection can be employed by
simply evaluating only the test image. For example, in FIG. 3, a
first image of a portion of mask M-2 that has only fine patterns
that are outside the imaging bandwidth of the inspector is captured
by the sensor 250 then processed by the processing circuitry 260.
Mask M-2 may have coarser patterns elsewhere that are within the
optical bandwidth of the inspector, but such patterns are not
relevant here. In this example, this first image of a defect-free
portion of the mask will ideally have no visible features since
these features are outside of the optical bandwidth of the
inspector. However, there may be some features visible in the image
due to imperfections in the tool itself. A simple method of
processing the first image would be to subject it to a filtering
step using knowledge of the imaging optics. The filtering step
could be implemented in the signal processing circuitry 260, or by
adding physical filter elements to appropriate portions of the
imaging optics.
[0040] Referring to FIG. 4A, a process 300 for operating on the
test image with the signal processing circuitry 260 is illustrated.
In step 302, a first image having an optical bandwidth consistent
with that of the inspection apparatus 200 is captured by the sensor
250. In step 304, the captured image is provided to the signal
processing circuitry 260. In step 306, a filtering function is
applied to the captured image by the signal processing circuitry
260. The filtering function is intended to remove the undesirable
portions of the image, i.e., remnants of the actual pattern, so
that only features of interest, such as defects, remain in the
resultant image. In step 308, the resultant image is evaluated by
the signal processing circuitry 260 to determine if any of the
remaining artifacts are, in fact, undesirable defects in the mask.
For example, if an EUV mask has lines which are not fully resolved
but not completely out of the optical band, and the pitch of the
lines is known, a filter can be developed to remove the frequencies
associated with the image of the lines, thereby leaving much of the
image of any defects.
[0041] FIG. 4B illustrates one embodiment of a process 320 for
implementing a filter in step 306. In step 322, the signal
processing circuitry makes a copy of the original image. In step
323, the signal processing circuitry processes the copy so as to
retain artifacts due to the processor but to remove possible
defects of interest. In step 324, the processed copy is subtracted
from the original, and returning to process 300, the result is
evaluated in step 308. In the absence of defects, this filter
should result in an image that is more uniform with fewer inspector
defects. Any remaining features in the resultant image are more
likely to reflect defects on the mask, which may or may not be
significant. One method of processing the copy of the image is to
replace every pixel with the average value of all the pixels in its
row before it is subtracted from the original. Alternatively, the
processed copy could have the column averages of the copy
subtracted from the original. In yet another alternative, the
previous two alternatives could be applied serially.
[0042] In general, inverse filtering can be applied to the original
image to generate a derived or "synthetic" image which enables the
processor to remove or otherwise process the original image to
thereby evaluate other image artifacts that are not in the original
image. Inverse filtering can be accomplished, for example, by
taking a Fourier Transform of the image.
[0043] Thus, a filter can be applied in the frequency domain to
remove the expected portions of the image (i.e., the pattern).
After filtering in the frequency domain and transforming back to
image space, the remaining artifacts can be evaluated as defects.
Alternatively, a filter can be applied in the image domain by
convolution. Further, physical filters can be implemented for
specific cases, but it is far easier and more flexible to implement
a processor control for an adjustable optical system.
[0044] FIG. 5A illustrates a simplified block diagram of one
embodiment of an optical inspection scheme. Illumination optics 410
direct radiation through mask 420 and into magnification optics
430, then through a filter 440, before the radiation is measured by
sensor 450. The sensor 450 is coupled to a signal processor 460,
which processes and evaluates the signals received from the sensor.
The filter 440 can be implemented in various ways as described
above, for example, as a specific optical filter, or as a
programmable adjustable optical filter. Using this scheme, the
image detected by sensor 450 is already filtered to remove the base
pattern such that all that remain are artifacts to be evaluated as
defects.
[0045] FIG. 5B illustrates another embodiment, similar to FIG. 5A,
except that the filter 441 is located after the sensor rather than
before. Using this scheme, the original image is first detected by
sensor 450, and then filtered to remove the base pattern such that
all that remain are artifacts to be evaluated as defects.
[0046] FIG. 5C illustrates yet another embodiment. In this
embodiment, the original image is detected by sensor 450 and
provided to the signal processor 461, which can implement any
programmed filter specifications to act on the original image, and
can be programmed to perform mathematical operations in the
frequency domain using the original image and the filtered image to
obtain a resultant image in the frequencies of interest.
[0047] Finally, FIG. 5D illustrates an embodiment in which the
original image is detected by sensor 450 and provided to the signal
processor 462. In this embodiment, various parameters of the
illumination optics 411 are adjustable, for example, the effective
bandwidth of the emitted radiation, the coherence of emitted
radiation, and the numerical aperture of the illumination optics.
Likewise, various parameters of the magnification optics 431 are
also adjustable, for example, the numerical aperture. Further, the
signal processor 462 can be programmed to implement filter
specifications to act on the original image; can also be programmed
to perform mathematical operations in the frequency domain using
the original image and the filtered image to obtain a resultant
image in the frequencies of interest; and further, can be
programmed to provide adjustment to the illumination optics 411 and
the magnification optics 431 in order to implement a specified
signal processing scheme.
[0048] As another example, one could design a filter that passes
everything that is not the desired image, e.g., as a low pass
filter and/or a high pass filter. Such a filtering function would
not return the desired image, but would return whatever artifacts
existed. The filtered image could then be subtracted from the
original image, and the artifacts would be removed, leaving only
the desired image.
[0049] In some cases, the base pattern of the mask is simply not
resolved by the inspection tool, especially in regions that are
periodic and have well defined diffraction orders that are outside
the numerical aperture of the inspection system, and yet the
inspection tool will still be able to detect small defects using
optical methods.
[0050] In the extreme example of a very small defect which is not
periodic, the inspection tool will image the defect with the same
size as its fundamental resolution spot, which would be large
compared to the desired structures on the mask. In this case, the
inspection tool can detect the defect directly on the test image
without the added noise or computational cost of subtracting the
reference.
[0051] In the case described above where one lithography step
produces fine, one-dimensional lines which are not fine enough to
be outside the optical bandwidth of the inspection tool, the
inspection tool can perform subtraction more efficiently if the
base pattern is removed, and yet still detect small defects. The
inspection tool can perform a variant of cell-to-cell inspection
where the cell displacement has a direction parallel to the lines
and a magnitude not constrained by mask. In this case, the
inspection tool can choose the magnitude of the displacement to
minimize spurious noise between the test and reference images, or
use multiple references to allow double arbitration, and, if the
chosen cell magnitude is not very much larger than defect image,
apply a matched filter designed with knowledge of that
magnitude.
[0052] As is true for in general for both inspection and
lithography, some illumination shapes are better than others, and
can affect both the signal and the noise. Highly coherent
illumination can minimize noise or interference from residual
resolution of patterns on the mask. As is well known in the art,
reducing coherence broadens the optical pass band, thus enhancing
the ability to sense patterns. Further, using highly coherent
illumination, inspection of masks with their patterns outside of
the optical band is very similar to inspection of blanks for pit
and bump defects, a technique that is well known in the art, and
many of the methods employed for optimizing blank inspections are
applicable here. For example, defects will often appear as phase
defects that have greater signal when the system focuses at
substantial offset from either the nearer substrate surface or the
nearer pattern surface.
[0053] Typically, a test image will be of a small portion of the
mask and the total area to be inspected. For example, with a square
pixel size of 55 nm, a total inspected square of 100 mm, and square
inspection portions of 1000 pixels, there will be approximately 3.3
million inspection portions. In embodiments which detect defects on
the test image without subtracting a reference image, it can be
both convenient and advantageous to subtract a synthetic reference
derived from the test image itself. A method well known in the art
for detecting phase defects on blank masks is to subtract the
column averages from columns of the test image portions and/or the
row averages from rows of that image. The synthetic reference could
include other estimates of the background. For example, a low-pass
linear filtered version of the image could be included in the
synthetic reference. Another possibility is equivalent to high-pass
filtering the test image portion before detection. Such filtering
would be advantageous in at least the situation where all of the
base pattern is outside the optical bandwidth of the imaging
system, but has spatial variations in the density of the pattern on
a scale much larger than the critical dimensions of the pattern but
small compared to, for example, the field of view of the imaging
system.
[0054] Interactions of the nearly or totally invisible base
pattern, the illumination shape, the illumination polarization, and
the detection analysis can affect the defect signal. Thus, there
can be tradeoffs between signal and noise in choosing illumination
for the mask and defects of interest. Preferred illuminations would
give larger signal-to-noise ratios.
[0055] There may be cases where the desired patterns on the mask
are nearly outside the optical bandwidth of the imaging system, but
not quite. In this case it would be preferred to restrict the
numerical aperture of the imaging to reduce noise in the test
images by putting the pattern more completely outside the optical
bandwidth of the system.
[0056] In the situation where the base pattern is outside the
optical bandwidth of the inspector, many standard inspection
algorithms will simply fail because they cannot align the test and
the reference images.
[0057] In the situation where there are one-dimensional lines, the
noise due to optical aberrations in the systems is substantially
reduced by using a time-delay-and-integration-based ("TDI") imaging
system. A TDI system, where the scan direction is in the X
dimension and the resulting sets of pixels are arrayed in the Y
dimension, the noise due to aberrations is lower if the X axis is
arranged to be parallel with the lines. Each pixel will average
over aberrations as a point scans in Y, resulting in a lower
expected aberration. Also each pixel in subtracted images will have
an identical aberration effect, since they are at the same Y, which
should cancel out in the subtraction. Generic cell-to-cell
subtractions would have greater noise due to different local
aberrations from one Y location to another.
[0058] While one or more implementations have been described by way
of example and in terms of the specific embodiments, it is to be
understood that one or more implementations are not limited to the
disclosed embodiments. To the contrary, it is intended to cover
various modifications and similar arrangements as would be apparent
to those skilled in the art. Therefore, the scope of the appended
claims should be accorded the broadest interpretation so as to
encompass all such modifications and similar arrangements.
* * * * *